Natural and social scientists have long had a practice of using asterisks ("stars") to indicate statistical significance in tables. I'll leave aside (for now) the question of whether this is a good practice or a bad one (and, for that matter, whether we ought to be doing statistical significance testing at all), and instead focus on the (mis)use of "stars."
Here's another example from a recent, well-respected law review. If we focus on the use of "stars," the table illustrates three potential issues:
(1) First, there are simply too many levels of them. Having separate numbers of asterisks for P < 0.2, 0.1, 0.05, and 0.01 is overkill. One or -- at most -- two asterisks is always enough; convention says to indicate P < .05 and/or .01, and stop.
(2) Second, there is no indication of the "tailedness" of the significance tests. This is important, if for no other reason than that it goes directly to how many stars appear next to each estimate.
(3) Finally, there's the issue of "stars" next to estimates for the constant term. Scholars routinely and unthinkingly include them (and, for that matter, P-values) in their tables even when they have no hypothesis about (or even any interest in) that term. Here, for example, the star on the constant in Model Two indicates that we can reject the hypothesis that the constant term is zero (i.e., that the probability of a lawsuit is 0.5 when all the other covariates are equal to zero) at P < 0.2. One's initial impulse to say "who cares?" is bolstered by the fact that such a condition implies that the hypothetical IPO amount was $0, which (I'd wager) is not an in-sample value.
I don't want to come across as completely negative. The authors managed to avoid a few other common mistakes when using "stars," including placing them next to the standard errors / t-statistics (rather than the parameters) or using a gaggle of arcane symbols (daggers, double-daggers, etc.) in place of asterisks. And, in general, this is a pretty good table; while the variable names are a bit cryptic, having variable descriptions in the table notes helps. I'll talk more about table-related matters in my next post.